Verifiably Human
Human-creation claims should identify their supporting process evidence, attributable declarations, and observation limits. Authority and accountability are evaluated separately. Anything the evidence does not resolve remains unknown.
Correction to Parts I–III
Authorized issuance is not verified origin. A person can retype or revise AI-produced text and issue a valid signed claim. Missing provenance means unknown origin, not synthetic origin. Capture, authorship, approval, and AI assistance can coexist. A dispute requires an adjudication process; it does not automatically invalidate a claim. An issuer’s stated time is not independent evidence of when a record existed.
Read the correction record →Four dimensions, reported separately
Process evidence
What activity was observed, under which threat model, and what remained outside observation.
Attributable declaration
What a particular issuer claimed about a particular artifact. A valid signature does not establish that the claim is true.
Accountability
Who investigates challenges, what evidence is considered, what consequences can follow, and how decisions are appealed.
Authority
Whether an actor was permitted to perform or publish a particular action or claim.
Open work
Define individual creation claims precisely enough that observed workflow, declared claim, and policy eligibility remain separate.
Compare which forms of process evidence survive deliberate substitution, and which legitimate workflows—including dictation, transcription, autocomplete, and assistive input—are wrongly rejected.
Measure attack acceptance, legitimate rejection, unresolved cases, privacy exposure, and collection burden. No benchmark result has been published.
Earlier essays remain available as historical arguments with visible editor’s notes: